My tests compared SQLite version 3 and the SQLcrypt™ based on it with MySQL version 4 and PostgreSQL version 8. (I have omitted PostgreSQL's numbers because version 8 is still in beta testing.)
The SQLite website contains a discussion of a similar set of benchmarks comparing SQLite version 2, MySQL version 3 and PostgreSQL version 7.
The two sets of tests show that, as the database engines evolve, their relative performance in various aspects change.
A test may invoke pathological behaviour in a specific database engine which causes it to perform poorly for that test. For SQLcrypt™, without instrumenting the engine, tests 3, 4 and 6 appear to do so. Implementation-wise, SQLcrypt™'s cryptographic operations happen at the storage layer and are oblivious to the workings of the SQL parsing and query processing layer. With "real world" applications, it is usually possible to write SQL that avoids observed pathological behaviour.
Discounting tests 3, 4 and 6, SQLcrypt™ is several times slower than MySQL in some tests, approximately as fast in some other tests, and outperforms MySQL in tests 2 and 14. Note that, with MySQL, no cryptography is applied to the data.